Stileex

The AI presence index of Malagasy tourism

How much room does Madagascar take in what AI assistants recommend to travelers? STAPI-MG measures that presence monthly, from 0 to 100, over a fixed basket of six travel intents.

In July 2026, the AI presence index of Malagasy tourism (STAPI-MG) stands at 14.7 out of 100: Madagascar is cited in 3 of the 6 travel intents in the basket. Where it is cited, the destination captures 29.3 points out of 100.

Stileex indexSTAPI-MG

Tourism AI presence, July 2026

14.7 / 100

Ceiling: 21.0 out of 100. The source rankings are truncated, so an absence is worth at most the last published rank.

The index decomposes into two factors

50 %

Breadth: 3 intents of 6

29.3 / 100

Depth where the destination is cited

14.7 / 100

STAPI-MG

Breadth multiplied by depth, divided by 100, gives exactly the index. It reads directly: a destination can be strong where it appears and still be barely visible because it appears nowhere else.

On its own continent, Madagascar ranks 17 of 22 destinations cited in the Africa ranking (leader: Tanzania). That ranking answers a conditional question and therefore stays out of the index.

The index basket, July 2026

One bar per travel intent: Madagascar's visibility in that ranking, from 0 to 100. A red bar is worth zero, the destination is not cited there. The dashed line is the index, the exact mean of the six bars.

The scale is pinned to 0-100: visibility reads as the share of AI-assistant answers citing the destination.

Travel intent Index term Published rank AI consensus Gap to the ranking leader
Safari 4 #16 of 25 2/4 94 points below Tanzania (98)
Wildlife 69 #6 of 25 4/4 25 points below Galapagos (94)
Ecotourism 15 #22 of 25 2/4 85 points below Costa Rica (100)
Beach 0 Not cited (at most 8) - 58 points below Maldives (58)
Adventure 0 Not cited (at most 17) - 71 points below Patagonia (71)
Honeymoon 0 Not cited (at most 13) - 79 points below Bali (79)

Each intent links the full ranking published by Charts Travel. 'Not cited' means AI assistants did not name Madagascar in that ranking this month: the term counts zero in the index, and the true value lies between zero and the ranking's last published index.

This index measures a perception: Madagascar's place in what AI assistants recommend to travelers, not its attendance. Real attendance is tracked on the tourism page, with the visitor arrivals at the borders published by the tourism ministry (MTA). The gap between the two is exactly what this index makes measurable.

Method

STAPI is the mean, over a fixed basket of six travel-intent rankings, of the destination's visibility in each: the share of AI-assistant answers naming it when a traveler asks that ranking's question. Probabilistic reading: the average probability that an AI assistant names Madagascar when a traveler asks one of the six basket questions.

An absence counts zero. That is the consistency condition of the measure: averaging only the rankings where the destination appears would make the index non-monotone, since losing its last citation in a ranking would RAISE it. The basket is fixed and versioned, and a month whose ranking could not be read is not published: an unread reading would otherwise count as an absence.

The source rankings are truncated to the leading destinations. An absence is therefore not a measured zero: the true value lies between zero and the ranking's last published index. The index is a floor on that account, and its ceiling, each absence valued at the last published index, is given with it.

Region rankings ('where to go in Africa') answer a conditional question: they stay out of the index, and the continental standing is published beside it. The Asia and Europe rankings are out of scope, Madagascar not belonging to those sets.

Method v1, basket v1. Every published point carries the method and basket versions that produced it: a later revision never rewrites history.

Where the readings come from

The readings come from the AI Travel Visibility Index, published by Charts Travel and measured by Epovest: consumer AI assistants answer a weekly panel of traveler questions, and the share of answers citing a destination gives its 0-100 visibility. The current month is provisional, revised at each new reading of the source, then frozen.

Stileex publishes the index and the Madagascar reading. The full rankings, the per-assistant detail and the granular history belong to the source.

API access

The month's index and the three figures that qualify it are freely available in JSON, no key. Every point carries the method and basket versions that produced it, and links the methodology: the value is citable and reproducible.

The index (latest month)
GET https://api.stileex.xyz/v1/series/mg/tourism-ai-presence/latest
The ceiling
GET https://api.stileex.xyz/v1/series/mg/tourism-ai-presence-ceiling/latest
The breadth
GET https://api.stileex.xyz/v1/series/mg/tourism-ai-presence-breadth/latest
The depth
GET https://api.stileex.xyz/v1/series/mg/tourism-ai-presence-depth/latest
The versioned methodology
GET https://api.stileex.xyz/v1/methodology/stapi
Series catalogue
GET https://api.stileex.xyz/v1/series/mg

The full monthly history, the longitudinal series that accumulates, is served on a free API key.

Frequently asked questions

What is the Tourism AI Presence Index (STAPI)?
A Stileex index published monthly, from 0 to 100: the average probability that an AI assistant names Madagascar when a traveler asks where to go, over a fixed basket of six travel intents (safari, wildlife, ecotourism, beach, adventure, honeymoon). It decomposes into breadth (in how many of those intents the destination appears) and depth (how strongly, where it appears).
Where does Madagascar stand?
The month's value sits at the top of this page, with its decomposition and the detail per travel intent. The table gives, for each intent, the term the index used, the published rank, the consensus across AI assistants and the gap to the destination leading that ranking.
Why does an absence count zero?
Because a destination AI assistants never cite in a travel intent has no visibility there, and that is exactly what the index must measure. Averaging only the rankings where the destination appears would make the measure non-monotone: losing its last citation would then raise the index.
Why publish a ceiling?
The source rankings are truncated to the leading destinations: an absence means 'outside the published destinations', not 'cited zero times'. The true value therefore lies between zero and the ranking's last published index. The index keeps the floor and publishes the ceiling with it, rather than claiming a precision it does not have.
Does the index measure real attendance?
No: it measures a perception, what AI assistants recommend. Real attendance is tracked on the Madagascar tourism page, with the monthly visitor arrivals at the borders published by the tourism ministry.